1.pace中采用的vgg16模型与之前使用的vgg16模型相差较多，无法进行对比
    image.png
    没有能够复现出paper中一模一样的结果，pace也存在随机性，但结果相差不大

    2.目前的效果
    m1:在class内进行聚类
    m2:对整个testing context进行聚类
    m3:昨天新提出的方法

    model精度 聚类个数 50 60 70 80 90 100 110 120 130 140 150 160 170 180
    mnist PACE(paper) 98.68% 0.603 0.319 0.088 0.070 0.245 0.320 0.419 0.487 0.551 0.611 0.653 0.703 0.732 0.764
    PACE(recur) 98.68% 0.641 0.347 0.109 0.054 0.233 0.310 0.411 0.480 0.545 0.606 0.649 0.695 0.728 0.761
    PACE(recur) 98.72% 1.280 1.280 1.280 1.280 1.280 1.280 1.280 1.280 0.517 0.566 0.613 0.659 0.688 0.721
    SRS 98.68% 1.303 1.537 1.374 1.183 1.274 1.209 1.077 1.025 1.193 1.014 0.942 0.974 0.855 0.836
    SRS 98.72% 1.745 1.367 1.330 1.578 1.335 1.106 1.272 0.768 0.944 1.043 0.845 0.779 0.783 1.023
    m2 98.72% 20 0.681 0.359 0.071 0.030 0.156 0.270 0.354 0.373 0.258 0.159 0.053 0.030 0.090 0.169
    12 1.280 0.333 0.149 0.060 0.144 0.249 0.363 0.440 0.511 0.566 0.613 0.022 0.495 0.377
    m3(1396) 98.72% 20 1.280 1.280 0.149 0.019 0.144 0.720 0.538 0.387 0.258 0.833 0.707 0.572 0.464 0.287
    m3(1396)_scale 98.72% 20 1.280 1.280 0.149 0.002 0.144 0.761 0.555 0.387 0.247 0.833 0.707 0.595 0.474 0.387
    m3(4000) 98.72% 20 0.761 0.415 1.661 1.220 0.894 0.720 0.555 0.359 0.258 0.848 0.707 1.204 1.073 0.955
    m3(4000)_scale 98.72% 20 1.280 1.280 0.169 0.014 0.169 0.290 0.371 0.447 0.511 0.138 0.045 0.583 0.485 0.918
    vgg16 PACE(recur) 87.41%
    		2.786 2.754 4.139 2.590 2.590 2.590 2.498 2.673 2.666 3.304 3.923 3.785 2.531 1.976
    SRS 87.41%
    		4.737 4.103 4.637 3.947 3.671 3.620 3.034 3.116 3.122 2.869 2.490 2.370 2.356 2.692
    m2 87.41% 15 3.096 2.410 1.299 0.231 0.230 0.709 0.878 0.090 0.282 0.176 0.077 0.011 0.237 0.858
    m3(1224) 87.41% 15 1.410 2.410 0.453 3.660 2.795 2.944 2.725 2.410 1.914 0.981 1.884 2.410 2.030 1.299
    m2(1224)_scale 87.41% 15 1.696 0.743 0.453 0.090 1.601 0.155 0.137 0.923 0.187 0.360 0.509 0.453 0.275 0.299
    m3(4000) 87.41% 15 3.737 4.359 4.311 3.866 3.140 4.410 5.428 5.592 4.204 3.839 3.410 2.410 2.030 1.854
    m3(4000)_scale 87.41% 15 0.345 0.726 0.086 1.334 0.368 1.479 0.137 0.633 0.487 0.267 0.590 1.340 1.414 1.354